Glück, Indurkhya / CHANGE IN LONGITUDINAL SALIENCE Assessing Changes in the Longitudinal Salience of Items Within Constructs

In longitudinal studies, marked changes in item salience may occur even if the internal consistency of a scale remains stable over time. Such changes in the pattern of response probabilities are relevant for the substantive interpretation of the scale scores. Also, there may be interindividual differences in item salience within time points. The article demonstrates how Rasch models can be used to study item salience within and across time points. Also, the authors want to demonstrate that even if a test or questionnaire does not fulfill the assumptions of the Rasch model, it can be used to obtain information on item functioning across participants and over time. The method is illustrated using data from the Baltimore Prevention Program. The use of adequate measurement instruments is an important issue in the design of longitudinal studies. These instruments are used periodically throughout the course of the study to obtain data on participants, giving rise to repeated measurements. Researchers usually use factor analytic methods to identify and group the items from the instrument to create substantively meaningful constructs (see, e.g., Nunnally, 1978). Generally, constructs are determined using data from the initial time point of the study, and then these constructs are checked for reliability and/or validity at each subsequent time point during analysis. Researchers have addressed methodological questions regarding the longitudinal stability of constructs using methods such as latent growth models (see Magnusson, Bergman, Rudinger, & Torestad, 1994). However, little attention has been paid to the longitudinal salience of the items that form these constructs, which may be an important aspect, especially whe